We examine the biases inherent to chemical abundance distributions when targets are selected from the red giant branch (RGB), using simulated giant branches created from isochrones. We find that even when stars are chosen from the entire colour range of RGB stars and over a broad range of magnitudes, the relative numbers of stars of different ages and metallicities, integrated over all stellar types, are not accurately represented in the giant branch sample. The result is that metallicity distribution functions derived from RGB star samples require a correction before they can be fit by chemical evolution models. We derive simple correction factors for over- and under-represented populations for the limiting cases of single-age populations with a broad range of metallicities and of continuous star formation at constant metallicity; an important general conclusion is that intermediate-age populations (~1-4Gyr) are over-represented in RGB samples. We apply our models to the case of the Large Magellanic Cloud bar and show that the observed metallicity distribution underestimates the true number of metal-poor stars by more than 25%; as a result, the inferred importance of gas flows in chemical evolution models could potentially be overestimated. The age- and metallicity-dependences of RGB lifetimes require careful modelling if they are not to lead to spurious conclusions about the chemical enrichment history of galaxies.

We present Keck/DEIMOS spectroscopy of individual stars in the relatively isolated Local Group dwarf galaxies Leo A, Aquarius, and the Sagittarius dwarf irregular galaxy. The three galaxies--but especially Leo A and Aquarius--share in common delayed star formation histories relative to many other isolated dwarf galaxies. The stars in all three galaxies are supported by dispersion. We found no evidence of stellar velocity structure, even for Aquarius, which has rotating HI gas. The velocity dispersions indicate that all three galaxies are dark matter-dominated, with dark-to-baryonic mass ratios ranging from $4.4^{+1.1}_{-0.8}$ (SagDIG) to $9.6^{+2.5}_{-1.8}$ (Aquarius). Leo A and SagDIG have lower stellar metallicities than Aquarius, and they also have higher gas fractions, both of which would be expected if Aquarius were farther along in its chemical evolution. The metallicity distribution of Leo A is inconsistent with a Closed or Leaky Box model of chemical evolution, suggesting that the galaxy was pre-enriched or acquired external gas during star formation. The metallicities of stars increased steadily for all three galaxies, but possibly at different rates. The [$\alpha$/Fe] ratios at a given [Fe/H] are lower than that of the Sculptor dwarf spheroidal galaxy, which indicates more extended star formation histories than Sculptor, consistent with photometrically derived star formation histories. Overall, the bulk kinematic and chemical properties for the late-forming dwarf galaxies do not diverge significantly from those of less delayed dwarf galaxies, including dwarf spheroidal galaxies.

We applied three statistical classification techniques - linear discriminant analysis (LDA), logistic regression and random forests - to three astronomical datasets associated with searches for interstellar masers. We compared the performance of these methods in identifying whether specific mid-infrared or millimetre continuum sources are likely to have associated interstellar masers. We also discuss the ease, or otherwise, with which the results of each classification technique can be interpreted. Non-parametric methods have the potential to make accurate predictions when there are complex relationships between critical parameters. We found that for the small datasets the parametric methods logistic regression and LDA performed best, for the largest dataset the non-parametric method of random forests performed with comparable accuracy to parametric techniques, rather than any significant improvement. This suggests that at least for the specific examples investigated here accuracy of the predictions obtained is not being limited by the use of parametric models. We also found that for LDA, transformation of the data to match a normal distribution in the input parameters led to big improvements in accuracy. The different classification techniques had significant overlap in their predictions, further astronomical observations will enable the accuracy of these predictions to be tested.